|Product Performed to Expectations:||8|
|Specifications were sufficient to design with:||10|
|Demo Software was of good quality:||8|
|Product was easy to use:||10|
|Support materials were available:||10|
|The price to performance ratio was good:||9|
|TotalScore:||55 / 60|
Molex 2.4Ghz/5Ghz Antenna Kit RoadTest
by Gough Lui - November 2018
In our modern wireless connected world, the role of the antenna cannot be understated. Literally, the eyes-and-ears of your wireless radio/modem, any radio amateur would be familiar with the adage that a good antenna is the “best amplifier”. Having a good antenna pays in multiple ways – increased signal strength reducing data errors, allowing higher speeds, reducing air-time wastage, increasing range or reducing transmit power requirements for the same range, increasing sensitivity to weak signals. With our appetite for even smaller devices, smaller antennas are often sought, some of which come with significant compromises such as reduced gain and irregular radiation patterns. As someone who is quite familiar with Wi-Fi and ISM-band communications – the small transmit power, high noise-floor, wide bandwidth and high data rates pose challenges to the whole wireless subsystem.
When rscasny put out a call for people to RoadTest some antennas from Molex, as a recipient of a Tektronix RSA306 spectrum analyser and a licensed radio amateur, I thought this would be a good educational exercise to embark on given the specialist nature of the product. Unlike my last two RoadTests, this one will be in a “monolithic” format as it is the most suitable format.
Before I continue any further, I must first make it clear that I am not a professional RF engineer, merely a radio amateur. While I am aware of many radio concepts, I don’t claim to possess a complete theoretical understanding of how antennas work or have the most appropriate equipment to test and characterise an antenna. Instead, it is my attempt to do so given the constraints of my budget, space, equipment and improvisation skills. In the process, I am hoping to gain some insight into these antennas as well as the limits of improvised characterisation that can be performed with limited facilities. The measured performance may not be completely representative of the performance of these particular Molex products – e.g. when applied in different designs, used with different equipment, in different environments, etc. Please keep this in mind as you read this RoadTest.
When it comes to product design, integration of the antenna should be a primary consideration to ensure the RF performance of the whole system. This includes consideration of the type of antenna to be employed, its positioning relative to the device, its ground plane requirements and any necessary matching or filtering components.
The most basic antenna is basically a “wire” – which on many designs can be accomplished as a PCB trace. This is often used with many low-cost consumer electronics devices, but has a number of drawbacks due to the physical size required for implementation and dependency on fabrication accuracy and material properties of the PCB with limited opportunities for tuning.
Moving up from this is the “chip” antenna which is often made of ceramic or plastic. This is a separate component, but being similar to SMD resistors, they can be automatically mounted during production. As these components are manufactured to specification, they should be relatively consistent and come with design documentation that eases integration into your design. Many of these chip antennas require additional filtering/matching inductors/capacitors, which provides an opportunity to tune the design even after manufacture.
The final type is a cabled/flex type antenna, which is basically a self-contained antenna design. By moving the antenna to one of these products, you basically provide a 50-ohm coaxial connection from your radio and the antenna does the rest. Attention still needs to be paid to the mounting of these antennas, and generally these antennas tend to be bulkier and more expensive, but this can simplify design dramatically. In the case of using a cabled antenna, there is also the potential to exchange antennas to meet different needs or cable out to an external antenna socket.
This RoadTest focuses on three Molex products. The first is the 206513-0001 which is a 2.4Ghz SMT Ceramic Antenna. This is similar to a regular chip antenna, measuring 3x3x4mm. The second is the 47948-0001 2.4Ghz SMT On-ground MID Chip Antenna, which seems to be a “plastic” version of the ceramic antenna, sharing very similar dimensions of 3.14x3.14x4mm. The final product is a 206995-0150 2.4/5Ghz PCB Antenna, which is a cabled self-adhesive ferrite-backed antenna suitable for mounting on metal, pre-terminated with 150mm of cable with an MHF connector and measuring 20.5x20.5x4.8mm.
But lets first see how the Molex products compare to those of the competition. Please remember – every market survey cannot be exhaustive and only compares a limited amount of on-paper specifications. Purchasers are advised to confirm specifications meet their needs prior to purchasing. Information is given in good faith for comparison purposes only.
The above table was compiled from all of the Chip/SMD type Wi-Fi antennas available on the Australian element14 site on 19th October 2018, comparing some key specifications including:
As can be seen, there are ample competitors in the market, but the Molex solutions appear to keep pace with the competition. They boast an above-average gain (although not the highest) and dimensions which are comparable with most other antennas with the exception of height which is taller than most. The pricing of the ceramic antenna is rather competitive, although the MID antenna appears to be more average. In all, it seems their solutions are competitive with the market.
Repeating the same survey but for cable/flex type products shows fewer competitors. Unfortunately, there is no pricing information for the ferrite-backed PCB antenna included in this RoadTest, but compared to the competition it seems to have a rather average gain. In terms of dimensions, it is quite a bit thicker and wider than most others, as many of the other products are flexible printed antennas that have virtually zero thickness. However, these antennas cannot be mounted on metallic surfaces – so the additional thickness of the ferrite layer and the PCB “square” construction may be the unique drawcard that distinguishes the Molex product from the remainder.
While Molex may be more well known for their connector products, it’s interesting to know that they have had a track record of providing custom antenna solutions and continue to do so. I think this aspect really shows when you start comparing the datasheets for Molex products compared to some of their competition – the Molex products have much more comprehensive data available including their test results over a range of frequencies for different ground-planes, mounting positions, etc. As a result, a designer can feel a lot more confident about their component choice based on having this data on hand and potentially reduce the amount of redesign required later on to optimise the design.
The antennas arrived in a padded yellow envelope inside a box.
Three antennas were included. The first is a self-adhesive cabled PCB antenna with ferrite layer, terminated in a 150mm cable with MHF connector and spacers.
The PCB trace design on the front side, if anyone is interested, looks as follows:
The second is a ceramic SMT antenna, mounted to their evaluation board with some rigid coax to an SMA connector.
The matching components are included on the board.
Closely looking at the pattern on the antenna, it seems there is some slight irregularities at the edges. I wonder if this might affect the performance of the antenna? The printed pattern forms a “squared” helix.
The antenna is shown next to a 0.5mm gradation scale to show the size.
The rear of the board is not populated, and forms a ground plane with many vias bonding it to the ground plane on top.
The final antenna is the on-ground MID antenna, which seems to be similar although formed on a plastic former.
Looking at this one, there seems to be some melting of the former during mounting of this antenna to the PCB. This probably doesn’t affect the performance as it’s on a non-driven pad, but it suggests that the soldering profile may be more critical for this part.
The matching components and the antenna is shown next to a 1mm and 0.5mm scale. This particular antenna requires two matching components as compared to the ceramic antenna which requires just one.
In order to attempt to determine the performance characteristics of the antennas, I attempted to measure the gain of the antenna by running indoor testing in my room with my Tektronix RSA306 spectrum analyser and a Mikrotik hAP mini as a signal source. Because of the crowdedness of the ISM band and the rather “delicate” nature of RF measurements, there were a number of considerations and decisions that were made to ensure the most “accurate” measurements given my limited facilities.
Testing was performed indoors, in my room, with the Mikrotik hAP mini as the signal source and the Tektronix RSA306 spectrum analyser as the receiver. The distance between the two was fixed at 2.50m to ensure that the signal would be in the “far-field” (i.e. not inductively coupled by near-field effects). The height of the receiving element is maintained at the same height as the access point. I also ensured that antennas were in the vertical polarity.
The Mikrotik hAP mini was chosen as the signal source as it has a number of useful abilities. One is to operate in the proprietary nv2 mode, which is a TDMA mode with 1ms cycle time ensuring a large number of “bursts” to measure. The second is the ability to adjust transmit power from 0-17dBm in 1dBm steps, which (to my knowledge) is loosely calibrated at the factory. To access the highest transmit powers requires ensuring the AP transmits only in the slowest modes (MCS 0). As a result, I chose to lock the unit to MCS 0 operation only. The final advantage is the ability to operate in “reduced bandwidth” modes – instead of the normal 20/40Mhz mode, it can operate as narrow as 5Mhz which increases power spectral density allowing the signal to better stand out from the noise.
Testing in the ISM band is a bit problematic without an anechoic chamber and Faraday cage because the noise floor from other competing signals will affect measurements, and radiated signal has the potential to interfere with services that you require. As I operate a Wi-Fi station in 40Mhz wide mode, I already occupy a good portion of the band. My neighbour likes Channel 6 in 20Mhz mode, so it already overlaps with my own AP to some degree. Finally, Channel 11 is useful to me – a weak distant AP provides my backup WAN connection, so I wouldn’t want to annihilate this either.
In Australia, we are permitted operation up to Channel 13 for WLAN services, which is shown in grey. I do have a limited number of Channel 13 neighbours, but they are so weak that they cannot be received in my room. As a result, I decided to centre my test carrier at 2477Mhz in 5Mhz wide mode – this avoids overlapping with my own services and avoids overlapping with the strongest neighbours in a section of the band that is completely free of overlapping from stations operating in the 1-6-11 channel scheme. Good for me, good for my neighbours!
Measurements were made on the Tektronix RSA306 in Signal Strength mode with the centre frequency and bandwidth matched to the transmitter. Measurements were made with 1024 samples averaged, each sample being of length 258ms (ensuring 257-258 bursts are averaged in each sample). This ensures that instantaneous variations of signal strength are not reflected in the measurements.
To guard against potential error in the output power from the hAP mini, which is not designed as a calibrated RF source, I decided to measure every power step from 0-17dBm on the hAP mini and perform a linear fit to average out spot-sample errors. Background noise levels were also measured to ensure that signals were not being adversely impacted (averaged level was no higher than -88dBm, providing over 20dB margin to the weakest measured signal).
A test of a single antenna hence took 9.5 hours to complete with analysis and write-up additional. Unfortunately, owing to differences in connectors and the way the Molex evaluation boards were assembled, a perfectly fair test was not accomplished. For commercial RP-SMA antennas, they were connected through an RP-SMA to N adapter to the RSA306. For the Molex antennas, an SMA to N adapter was used to the RSA306 – slight differences in the adapters may have affected the results slightly (approximately +/- 1dB). Additional losses are introduced in the connectors, short length of rigid coax and PCB matching network which are not replicated in the test of the commercial RP-SMA antennas.
One complication was that the ferrite-backed antenna was supplied with an MHF connector which I could not adapt into my test rig. As a result, testing could only proceed after tests of the other antennas were completed and the “tail” was desoldered from the ceramic evaluation board and soldered to the antenna. The first time resulted in a short, so the soldering was later re-done.
My first test was just with a 2dBi whip to see if the test is repeatable and generates the results expected. The tests showed that the measurements of power at each spot was accurate to about 1dB (as expected based on datasheet performance). Both fitted lines were visually close, indicating good repeatability, and path loss was estimated to be ~48.5dB which is close to the FSPL of 48.28dB. This seems to confirm the validity of the experiment and the linearity of the power output control on the hAP mini.
I then tested the three Molex antennas to rather confusing results. All of the antennas achieved lower gains than the 2dBi whip which was contrary to the datasheet claims. This could have been due to the additional losses in the PCB, connecting rigid coax and connectors, but I didn’t expect to read gains of around -4 to -9dBi. As a result, I would take this result with a grain of salt. Maybe this could have been affected by multipath propagation or by poor gain at 2477Mhz which is at the upper end of the band.
In my confusion, I decided to test absolutely every RP-SMA antenna I had in my possession. Rather surprisingly, a 9dBi card Yagi measured about 6dBi which was understandable due to having about 1m of connecting coax, but an 16dBi Yagi measured -5.4dBi which suggests that it has much lower gain despite having the same length of connecting coax. Maybe that coax was quite poor quality and being a low-cost unit from China, its specifications could be quite loose. Comparing two TP-Link antennas of 4dBi and 9dBi sees the 9dBi achieving lower gain than the 4dBi, both of which are below that of the 2dBi whip. These results were not what I expected, but seem to confirm that antenna specifications are tricky, as higher gain antennas typically have lower operating bandwidths and “peak” gain can be quite different from average gain.
The summary of the computed dBi gains based on linear fitting with constrained gradient of 1 is as shown on the graph above. That being said, I would take the results with a grain of salt – I am not an antenna testing laboratory.
The raw measured signal strength data is as follows:
Dissatisfied with my results above, one of my hypotheses was that the reduced gain exhibited was due to operation far from the optimal frequency. On the assumption that most antennas are tuned to achieve maximum performance “mid-band”, I tried to perform some spot testing at 2452Mhz (the closest channel step the Mikrotik would allow) with the Mikrotik set to 17dBm transmit power to overcome the slightly higher noise floor at this frequency.
Unfortunately, instead of improving the comparative gain results between the 2dBi whip antenna, the performance seemed to get worse for all tested Molex antennas. The reasoning seems unclear – maybe it was losses in connectors, cables, mounting on PCB/matching or a particularly good whip antenna. But the ferrite-backed antenna seems to fare the best at this frequency, so this change does suggest gain is frequency dependent (but we know that already – from the datasheet).
A look at the spectra does, however, show that the antenna seems to operate more happily at 2452Mhz – I’m not sure if the “waviness” in the signal in the tests at 2477Mhz was due to antenna mismatch at the transmitter or receiver, but the signal having a “flat top” is what we expect to see.
Rather than rely on only an attempt to characterise the antenna under “artificial” conditions, I also decided to test the antennas in “real life” on a production network using an external Wi-Fi card and iperf3 to determine throughput at various places around the house. However, even this has some limitations, which are discussed in the methodology section that follows.
To interface the Molex antennas, an Amphenol RF 132168RP RP-SMA to SMA adapter was used to connect the antenna to the RP-SMA port of the TP-Link Archer T2UHP 802.11ac USB Wireless adapter. This particular model of adapter was chosen as it is a single-spatial stream adapter, hence able to operate with just one antenna, and was shown by my earlier teardown not to contain any vestigial internal antennas that could render the results meaningless. As the adapter has a recessed RP-SMA port, the casing was disassembled for the duration of the test.
Throughput testing was performed using iPerf3 under Windows 10, highly regarded as the most accurate speed testing tool. The server was hosted on a computer connected to the router via Gigabit Ethernet. The client was running on my Lenovo E431 laptop. Testing was commanded via remote control over the 5Ghz network using the laptop’s internal wireless card, whereas the iPerf3 data traversed the 2.4Ghz network (separate SSID) using the USB wireless card through the use of binding options in the iPerf3 command line. Tests were run in a single stream, TCP mode for 60 seconds in one direction at a time.
Five locations were chosen around the house for testing, with 10 tests in each direction being run and the best 3 results recorded. Testing hence required around 7 hours in total. As the network used for testing was my “regular” home production network, there was a small amount of traffic which could have affected the results. But it was chosen to use this network for two reasons – a separate network would not negate the problem of interference and the limited bandwidth of the 2.4Ghz spectrum precluded the operation of a 40Mhz network (which is most demanding for the antenna in terms of bandwidth requirement) in a non-overlapping mode. As a result, increasing test length and repetition was seen as a way to guard against variations.
In the AP to client direction, the test results showed roughly similar results between antennas and a reference 2dBi whip. The key difference was at the study and property boundary, where the weaker signals were encountered and the Molex ferrite-backed antenna produced notably lower throughput. Other than that, the results were fairly close and likely within the margin of error.
In the reverse direction, aside from the strong signal in the downstairs bedroom, there seemed to be a clear preference for the 2dBi whip. The reason for this isn’t clear, but could potentially signify a radiation pattern difference or lower antenna gain. Why this was not manifested in the other direction could be due to two reasons – the AP is a Mikrotik hAP ac which has a massive 29dBm transmit power thus it could probably overcome reception impediments with brute power and the wireless card may not be bothered if the signal still has enough SNR even if the absolute signal strength is slightly lower. In the return direction, the power generated from the T2UHP is around 17-20dBm which is a bit less, thus the signal at the AP may not achieve sufficient SNR to maintain higher bitrates. This does suggest that the gain of the Molex antennas in the tested configuration is probably lower – the reasoning is not clear though.
The raw results are given below.
Antennas are a critical component in wireless systems and their proper design and integration is paramount to the performance of the radio/modem. To this end, Molex offers a variety of antennas in their portfolio that appear to be competitive with the market, with a number of specific innovative products (e.g. ferrite backed antenna for on-metal mounting, multi-band antennas, MID type antenna). Molex also offers custom antenna designs and their product specifications are particularly comprehensive compared to the competition, which should simplify antenna selection and integration.
The particular “chip” type antennas provided seem to be a helical design deposited onto a former of either ceramic or plastic material which is taller than most chip antennas. The provided ceramic antenna seemed to have slight imperfections in the printed pattern, with the plastic unit seeming to have suffered during the population process onto the evaluation board. The ferrite backed PCB antenna seems reminiscent of regular PCB antennas, although square in shape and slightly thicker.
However, it seems that RF characterisation is probably something that is “best left to the professionals”. While I did have a simple indoor rig to test the gain of the antennas which seemed to produce the expected result (based on free-space path loss) on the reference dipole antenna in a repeatable manner, tests of other antennas resulted in readings that were unusual. This can be attributed to a number of issues – losses in connectors/adapters and cables, gain-frequency variations (i.e. maximum gain is only at one very specific frequency), radiation-pattern variations, signal reflections, construction of the evaluation boards themselves and potentially an “uncalibrated” RF source. As a result, the Molex antenna results were rather unflattering at 2477Mhz and even spot-checks at 2452Mhz did not produce values which I would consider expected. In their defense, measuring other commercially-made antennas and comparing them with their “claimed” gain values resulted in equally unpredictable results. As a result, while my testing suggests that the Molex antenna evaluation boards had a notably lower gain than a dipole antenna, this may be specific to my test set-up and is inconclusive.
Attempting to firm-up the results by comparing the antenna performance over a more-demanding 40Mhz-wide Wi-Fi carrier using iperf3 to measure throughput resulted in results that showed similar performance from AP to client with a few minor exceptions in the Ferrite antenna which seemed to have some difficulty at lower signal levels. In the reverse direction from client to AP, it seems that the reference 2dBi dipole antenna had a slight advantage in around half of the locations. Due to the load and noise-dependency of the tests, such slight differences are not seen to be particularly significant as they could arise due to “localised” interference events, slight differences in positioning causing multi-path, etc.
As a result, I am unable to make any definite conclusions on the performance of the Molex antennas. While you can run some basic tests, the latter test suggests that the performance is comparable to a regular 2dBi dipole whip despite the appearance of a lower gain in the testing with a spectrum analyser without being able to provide a clear reasoning or understanding of the margin of difference.
Thanks to Molex and element14 for providing the antennas for this RoadTest. It’s regrettable that my proposed experiments were not able to demonstrate any conclusive results, but I would have to say that I have invested a lot more time than I thought I would and learnt a bit along the way.